Abstarct: In this thesis, a codec system is designed for high resolution textual images baxsed on super resolution. Spatial resolution is very important in textual images, but this feature will increase the storage size of the images. Due to the limitation of digital memories and large size of these images, textual images compression seems necessary, But there are some problems; 1) Compression can cause destructive effects on textual images. 2) In methods like JPEG and JPEG2000, compression ratio or output bit rate, is limited. The main idea in this thesis is to provide a good way to further compression ratio and also reduce the damaging effects on compressed image. To achieve greater compression ratio we use the idea of reducing the size of textual images. It should be noted that the resizing may has a effect on image quality and may cause degradation in image quality. So a procedure must be chosen that can increase the size of textual images and improve the devastating effects on images. In this result, the idea of super resolution combined with compression methods has been used. The methods of super resolution baxsed on interpolation and baxsed on learning in the step of reconstruction, have been utilized. In the method baxsed on interpolation, input low resolution image is divided into three laxyers and then each laxyer baxsed on its importance, has been magnified with a special method. At last magnified laxyers have been combined to each other and have formed the final high resolution image. In the method baxsed on learning, at first a suitable data baxse of high resolution and low resolution textual images, have been caused. Fore making dictionary, 3*3 and 5*5 patches have been used. The dictionaries have been baxsed on feature and four high pass filters for having high frequency components of images have been used to make dictionaries. Dictionaries have been learned for 50000 and 100000 atoms. The created databaxse including real document images with high and middle resolution. For an image that was scanning with spatial resolution 600dpi, JPEG and JPEG2000 methods can compression up to 0.07 and 0.015 bit rate respectively; proposed method for this image can compression up to 0.004 bit rate for both methods. The results baxsed on Optical Character Recognition (OCR), Mean Opinion Score (MOS) and Peak Signal-to-Noise Ratio (PSNR) criterions, have been evaluated and compared with each other. Proposed method baxsed on OCR and MOS criterions that are important in textual images, has better results than all other methods but the PSNR for proposed method was not good.